When creating tarballs using the standard tar -cvjf command, the archive preserves the full directory structure of the source files. For example:

tar -cvjf backup.tar.bz2 /home/user/projects/myapp

When extracted, this creates the full path /home/user/projects/myapp in the destination directory. This is often undesirable when you just want the contents of myapp without the parent directory hierarchy.

The most efficient solution is to use tar's -C (change directory) option combined with relative paths:

tar -cvjf clean_backup.tar.bz2 -C /home/user/projects myapp

This command:

1. Changes to /home/user/projects before processing
2. Adds only the myapp directory contents
3. Creates an archive that will extract directly to ./myapp

For more complex path manipulation, GNU tar offers the --transform option:

tar -cvjf transformed.tar.bz2 \
    --transform 's,^home/user/projects/myapp/,,' \
    /home/user/projects/myapp

This uses a sed-like expression to strip the prefix from paths in the archive.

Here's a complete example showing both creation and verification:

# Create archive without parent paths
tar -cvjf app_release.tar.bz2 -C /builds/v1.2.3 app

# Verify contents
tar -tvf app_release.tar.bz2 | head -5
# Should show paths starting with 'app/' not '/builds/v1.2.3/app/'

# Extract test
mkdir test_extract && cd test_extract
tar -xvjf ../app_release.tar.bz2
ls
# Should show 'app' directory directly in test_extract

This technique is particularly valuable when:

• Creating deployment packages that need predictable extraction paths
• Building Docker images where layer paths matter
• Sharing code where recipients expect flat directory structures
• Automating builds where absolute paths would break portability

---

When creating tarballs of specific directories, you might have noticed that the archive preserves the full directory path structure by default. This means when you run:


tar -cvjf archive.tar.bz2 /path/to/folder/source


The resulting tarball will contain the entire "/path/to/folder/" hierarchy when extracted. This is often undesirable because:

1. It forces extracted files into specific directory locations
2. Makes the archive less portable between systems
3. Requires additional steps to move files after extraction

The most reliable method uses tar's -C (capital C) option to change the working directory before processing files:


tar -cvjf destination.tar.bz2 -C /path/to/folder source


This command:
1. First changes to /path/to/folder (-C option)
2. Then archives just the "source" directory
3. Preserves relative paths within source

For more complex path modifications, GNU tar offers the --transform option:


tar -cvjf destination.tar.bz2 --transform 's,^path/to/folder/,,' /path/to/folder/source


The transform flag uses sed-style replacement patterns to strip paths. This is particularly useful when:
- You need to modify multiple path components
- Want to replace paths rather than just remove them
- Need to handle complex archive restructuring

Here are some real-world scenarios with solutions:

Example 1: Archive a web application's public directory


# Bad (preserves full path):
tar -czvf app.tar.gz /var/www/html/public

# Good (clean paths):
tar -czvf app.tar.gz -C /var/www/html public


Example 2: Create a backup of user home directories


# Without parent paths:
tar -cjf users_backup.tar.bz2 -C /home user1 user2 user3

# With directory flattening:
tar -cjf flat_backup.tar.bz2 --transform 's,^home/,,' /home/*


• Always test your archives with -t (list) before extraction: tar -tvf archive.tar.gz
• Combine these techniques with exclude patterns for more control
• For maximum portability, use relative paths rather than absolute
• Remember that -C affects all subsequent path arguments

For ultimate control over paths, you can combine find with tar:


find /path/to/folder/source -printf "%P\\n" | \\
tar -cvjf destination.tar.bz2 --no-recursion -C /path/to/folder/source -T -


This approach:
1. Uses find to list relative paths (%P)
2. Pipes them to tar's -T (files-from) option
3. -C ensures proper base directory
4. --no-recursion prevents automatic directory traversal